There are multiple open bugs about NFS for Debian unstable right now.
One description of the problem says that the breakage is triggered by nfs-kernel-server transition to rpcbind - and rpcbind and portmap cannot bind the same port at the same time. However, rpcbind package does not have an init.d script at all, although it seemed to work on my client when started manually (my server is broken as well, so I didn't check it fully).
However, I do not wish for a manually started service on a server that may be rebooted at any time - and I don't wish for a hacky script written by me or some random guy.
So, I'm looking for the "official" solution fixing NFS usage for Debian that doesn't involve manual hacking. The proper solution will obviously take some time, but a good interim solution what is needed by every Debian NFS user world wide.